"Meta is a very demanding customer": Volozh on Nebius deals and industry bottlenecks

Multibillion-dollar contracts with bigtech companies are not the ultimate goal of Nebius, but a tool for building a large-scale cloud business, the company's head Arkady Volozh said in an interview with Business Insider. According to Volozh, the deal with Meta is a financial foundation that provides a guaranteed cash flow for the development of high-margin core services. The co-founder of Yandex also spoke about barriers in the AI market, "pie economy" and solving the problem of idle chips.

The industry faces four fundamental constraints that have become radically more acute over the past year: capacity, capital, chips and customers, Volozh said. First, infrastructure is not keeping up with demand - including a shortage of basic power equipment. "The physical world is simply not able to build data centers fast enough," he stressed. Second, the entry threshold has risen: to capture 10% of the AI infrastructure market requires capital expenditures of about $400 billion. Third, the deficit in the semiconductor market has shifted from processors to memory chips. Finally, customer demand for AI technologies now has a solid economic basis and is far ahead of the supply that the industry is capable of creating, the Nebius CEO said.

Nebius has no plans to resell other people's computing power: the company's strategy is to become one of the market leaders in cloud services. "Consider Nebius as the fourth hyperscaler. You can't achieve that by acting as a hardware wholesaler. You have to own the entire stack (full technology cycle. - Oninvest)," Volozh explained. The company designs the equipment itself and controls the value chain. "Owning the platform allows us to efficiently distribute capacity and create exactly what customers need. This is how we attract corporate clients," he emphasized.

Nebius' competitive advantage is built on controlling all five levels of AI infrastructure creation, from laying the foundations of data centers to its own digital solutions. "We control our cost structure from concrete to software. This is how we deliver better economics," Volozh stated. The company saves 15-20% on self-assembly of server racks (second layer). Wholesale capacity sales to large players (the third layer) gives scale, while the main profit comes from the top layers - selling cloud and AI services to corporate clients.

A management team with years of experience at Yandex allows Nebius to handle infrastructure challenges without contractors, from land buyouts to power contracting. "We understand the day-to-day complexities of rapidly building data centers. <...> We do not outsource these complex tasks," Volozh stated. Direct work with the authorities also helps the company to stay on schedule.

Due to software imperfections, expensive chips are often underutilized, generating losses. "Shadow GPUs are idle computing power. Customers pay for them but can't use them efficiently" as competing platforms are not configured to maximize utilization, Volozh argues. Nebius solves this problem with a cloud environment designed for AI developers. The company's customers always know the exact amount of their available capacity. "It's about more than just avoiding downtime. It's about partnership," he explained.

Deals like the $27 billion contract with Meta are not the end goal. Meta is a "very sophisticated and demanding customer," but Nebius' core long-term business is selling high-margin AI services to a broad market. "Large contracts with Meta and Microsoft are fuel: they allow us to build the core business faster," Volozh summarized.

This article was AI-translated and verified by a human editor